DIMENSIONAL INFORMATION FACILITATES THE UTILIZATION OF CONFIGURAL INFORMATION - A TEST OF THE CASTELLAN-EDGELL AND THE GLUCK-BOWER MODELS

In 4 experiments involving learning in a probabilistic environment (a nonmetric multiple-cue probability learning environment) the authors f ound that utilization of relevant configural information was facilitat ed when any relevant dimensions were part of the relevant pattern. Rel evant configural information in an environment with no relevant dimens ions was as highly utilized as when the pattern contained the relevant dimension. With two relevant dimensions, the dimension within the rel evant pattern was more highly utilized than the dimension not in the r elevant pattern. Both N. J. Castellan and S. E. Edgell's (1973) hypoth esis testing model and M. A. Cluck and G. H. Bower's (1988a) connectio nist model were unable to fit several aspects of the findings.